Multi-scale feature data and landscape analysis toolkit for predictive soil mapping
Spatial scales of soil forming processes are often different from the spatial resolutions of the environmental covariates used to predict soil classes or soil properties. Mixing environmental covariates of different spatial resolutions, especially when they do not match the soil forming process scales, can produce incorrect results in predictive soil mapping. An open source multi-scale or resolution feature landscape analysis toolkit has been implemented to aid users in freely choosing desired scales or resolutions for predictive soil mapping sensitivity analysis and feature selection at process scales of interest. The multi-scale tool imposes spatial resolution by applying a series of mean filter neighbourhood sizes to an input elevation raster. For each user-specified resolution a series of derivatives are outputed, grouped into local, regional, and segmentation levels. The toolkit integrates open source software from GRASS, SAGA, R, GDAL, QGIS etc., and is packaged and distributed using a virtual machine image with the Ubuntu version of the Linux operating system. We describe the toolkit design, and demonstrate its use with a simple example, mapping soil A-horizon thickness on hillslopes using downslope index and conventional slope gradient. The predicted soil thickness distribution on hill slopes was significantly correlated with field data.
|Conference||5th Global Workshop on Digital Soil Mapping|
Geng, X., Burcher, R., Kroetsch, D., & Mitchell, S. (2012). Multi-scale feature data and landscape analysis toolkit for predictive soil mapping. Presented at the 5th Global Workshop on Digital Soil Mapping.